New techniques for early detection of lung cancer

There is a resurgence of interest in lung cancer screening, as new technologies allow for a more precise definition of which people are at risk, enabling early detection to obtain stage shift and allow appropriate treatment at the earliest possible stage 1. The lack of mortality reduction previously shown in randomised clinical trials (RCT) for lung cancer screening and recent analysis of the long-term outcome have led to hefty debates concerning whether itiscurrently appropriate to reconsider lung cancer screening in the general population 2–8.

Lung cancer morbidity is high and the cure rate is <15% 9. The majority of lung cancer patients are diagnosed at an advanced stage. Surgery provides thebest chance for cure if a tumour can be completely resected in the absence of lymph node and distant metastases. Even after curative resection, many patients remain at risk for distant metastasis, local recurrence and the development of subsequent primaries 10–13. Smoking-related problems, such as chronic obstructive pulmonary disease (COPD) and poor cardiovascular status, can limit the ability to apply thebest treatment strategies because of unacceptable morbidity and mortality. Smoking behaviour andcancer susceptibility in certain individuals also explains why many will develop mutifocal cancers, a phenomenon known as field carcinogenesis 13–16. The rate of second primaries is in the order of 1–2% 15. An earlier study has shown a cumulative incidence for second primary cancer of up to 4% per year 13. However, little is known about the natural history of cancer, i.e. carcinogenesis. Three pathways of preneoplastic lesions are currently recognised before lesions gradually progress towards squamous cell cancer (SCC), adenocarcinoma and neuroendocrine tumours 17–21. Microinvasive SCC gradually develops from normal bronchial epithelium in a period of up to 10 yrs. SCC may be multifocal due to field carcinogenesis leading to the development of multiple synchronous or metachronous primaries 21. Research to delay or reverse the gradual progression towards malignancy, e.g. chemoprevention strategies and molecular genetic studies, are beingperformed in many centres 1, 22, 23. As inhaled carcinogens can lead to cumulative genetic damage in the bronchial epithelium of (ex)smokers, biomolecular studies to understand the sequence of genetic changes are important to identify those at greatest risk 24, 25. The morphological changes seen through the various pre-invasive stages have been shown to be accompanied by more subtle genetic changes that are below the detection threshold of microscopic examination 24–29. Patches containing 200–600 cells have been shown to harbour molecular changes, e.g. loss of heterozygosity or microsatellite alterations 28, 29. Carcinoma in situ (CIS) consists of4–38 cell layers and the majority of these lesions areabout five cell layers thick. CIS lesions consist of noncircumferential patches, of several mm in magnitude, of atypical cells in the bronchial mucosa 18, 20. It is therefore obvious why conventional diagnostic tests such as chest radiographs, computed tomography (CT) scan and fibreoptic bronchoscopy (FB) have been insensitive for early detection and are not effective for early detection. Potentially malignant cellclones are far below the detection threshold of current diagnostic tests. It is paramount to develop more accurate and sensitive diagnostic tools for early detection and treatment.

Surgical resection and complete mediastinal lymph node dissection is the standard treatment of lung cancer. However, it remains a relatively morbid and wasteful treatment method for patients with central early stage cancer without nodal involvement, in whom superficial lesions are only several cell layers thick. As many patients may suffer from smoking-related diseases, less morbid and tissue conserving therapeutic approaches are indispensable elements fora cost-effective screening programme 30,31. Various aspects of new emerging technologies for early detection and treatment in lung cancer, mainly focusing on early SCC, will be discussed in this review.

The target population

Many Japanese centres have been conducting sputum cytology (and later on the use of low-dose spiral CT has been added to this programme) screening for decades 32, 33. Data have been consistent andin accordance with the Mayo screening study, showing that stage shift could be obtained. More casesof stage I–II squamous cancer amenable for surgical resection have been detected, leading to a higher resection and survival rate 34–37. Individuals with early stage SCC in their central airways seemed to benefit most because they had sputum-positive but radiographically occult lung cancer (ROLC). This seems to be in contrast to subjects with sputum-positive adenocarcinoma, who were shown to have apoor survival, as >82% of them were already in stageIIIA–IV 38. Data from randomised clinical trials performed by the three institutions (Mayo Clinic, Rochester, MN, USA; Johns Hopkins, Baltimore MD, USA; and Memorial Sloan, Kettering, NY, USA) showed that in the cohort of subjects with SCC,more early cancers were detected and survival after resection was significantly better compared to those who refused surgery, 63–76% versus <19%, respectively 4, 39. Therefore, despite many controversies, early detection of SCC appears to be ofsome benefit if combined with surgical resection.

Extended follow-up data reconfirmed earlier findings that no mortality benefit exists, despite the significant survival advantage in resected early stage (squamous cell) cancer cohorts versus those who refused surgery 2, 4, 6, 39. Lung cancer proved not to be the major cause of death in the individuals screened 4, 40. The mortality of smokers from all causes of death is about three-fold more than nonsmokers. Mortality due to lung cancer was ultimately 7% in both arms of the Mayo study and SCC formed a minor cohort. Many individuals died due to other reasons, including ischemic heart disease and COPD. Case survival differences in the absence ofmortality reduction have raised extensive debates about the real value of screening 4, 6–8, 40, 41. Overdiagnosis has been raised as an important problem of detecting relatively indolent cancers that would never have become clinically relevant. Some individuals are destined not to die from lung cancer per se, despite early detection and treatment. In the control arm of a screening study, the presence of lung cancer may remain undetected 41. Two per cent oflung cancer cases were missed out of the 153 primaries found in a post-mortem study 42 and deaths in the screened arm from other causes may have been falsely attributed to lung cancer. Healthcare workers and individuals at risk strongly perceive that early detection and treatment can save lives, because lung cancer is assumed to be a fatal disease 4, 6, 40. The inherent risks for mortality due to diseases such as COPD and cardiovascular diseases are competitive tolung cancer death as the ultimate cause. In retrospect, early lung cancer treatment may be regarded astreating a pseudo-disease. Screening may also lead to additional morbidity and mortality, which is not always expressed correctly in the number of disease-specific mortalities. All these factors have led to different interpretations and controversies regarding the potential value of lung cancer screening 40, 41.

As it is currently still impossible to accurately predict the cause of death for each individual upfront, it is ethically difficult to refuse treatment in case earlylung cancer is diagnosed. The lack of mortality benefit in previous RCTs, the issues of overdiagnosis and pseudo-disease may all be theoretically interesting, but they offer no solution in clinical practice 6. In previous decades, advanced disease at presentation has been named as the major reason for the dismal cure rate of lung cancer patients. Despite all theoretical arguments about overdiagnosis and pseudo-disease raised by the epidemiologists, a strategy of no treatment in a lung cancer screening trial will be unacceptable for clinicians.

The population cohort of (ex)smokers who have a 10% chance of developing lung cancer is very large. For peripheral early stage adenocarcinoma, a lobectomy and mediastinal lymph node dissection is the current standard of treatment 43, 44. In contrast, intraluminal bronchoscopic treatment (IBT) provides an alternative for surgery for patients with central type early stage SCC 30, 31, 45. The efficacy of thisminimally invasive technique allows for the treatment of poor surgical candidates who were previously denied any screening because they were considered inoperable.

It is important to realise that forced expiratory volume in one second (FEV₁), expressed as % predicted, is prognostic for death from ischemic heart disease 46. COPD with an FEV₁ <70% pred showed a 10-yr cumulative percentage for lung cancer death of8.8 versus 2% for those with an FEV₁ >85% pred47. The difference in survival at 10 yrs is significant, 74 versus 91.1%, respectively. The higher relative risk for lung cancer in COPD smokers was significantly and proportionally related to the degree of airway obstruction, and the presence of ventilatory impairment was associated with a 6.44-times greater risk for lung cancer 48. Paradoxically, screening of this population cohort may fuel more controversies regarding treatment of pseudo-diseases. Fortunately, IBT offers a less morbid treatment option to preserve quality of life, while it may significantly reduce the likelihood of dying from lung cancer 30, 31. Especially in this cohort of individuals, long-term local control to delay lung cancer death and not necessarily cure, may suffice to preserve quality of life. Therefore, the application of less invasive and morbid treatment methods has to beput into proper perspective.

Approximately 90% of (ex)smokers will not die because of lung cancer, therefore, a highly specific diagnostic test may allow the identification of the∼10% cohort who have the greatest risk. Molecular epidemiology studying cancer susceptibility, e.g. CYP1A1, CYP2D6 and GSTM polymorphisms looking for an individual's capacity to detoxify carcinogens, has shown that the presence of null GSTM1 and mutant CYP1A1 genotypes were associated with a relative risk of 41 to develop lung cancer 49. However, data have not been very consistent and a recent review by Kiyohara et al. 50 about genetic polymorphisms and lung cancer susceptibility has addressed this conflicting issue. Their meta-analysis seemed to show only myeloperoxidase and microsomal epoxide hydrolase exon 4 polymorphisms to be significantly associated with lung cancer risk.

The combination of accurate diagnostic tests and minimally invasive treatment are crucial elements forcost-effective screening. Screening-related anxiety, morbidity and mortality may increase burden and costs. The efficacy of minimal invasive treatment, i.e. intraluminal bronchoscopic treatment, as an alternative to surgery, warrants the inclusion of the cohortpopulation with severe COPD and poor general health, despite their inherent higher risk of dying from nonlung cancer-related diseases 30, 45. Itremains to be seen whether genetic susceptibility studies may help to achieve a more economical screening programme, instead of scrutinising all (ex)heavy smokers with or without pulmonary function impairment.

Sputum cytology screening

To detect early stage lung cancer, sputum cytology screening (SCS) and chest radiography (and later on low-dose spiral CT) have been performed in many centres 32, 33. Epidemiologists, cytopathologists, molecular biologists, radiologists and surgical oncologists have been involved in the process to identify andtreat those with early lung cancer. Individuals with suspicious sputum cytology (severe atypia or malignant cells) have been subjected to FB and further radiological staging 51. Sputum cytology mainlydetects SCC in the central airway. Treatments offered have been both surgical and bronchoscopic, e.g. photodynamic therapy 30, 32–38.

Positive sputum cytology for adenocarcinoma proved to be stage ≥IIIA in 82.4% of cases 37. Hence, low-dose spiral CT has been used to detect early stage parenchymal lesions. Data so far have shown that it is possible to obtain a significant stage shift for the detection of stage I/II adenocarcinoma in>80% of cases 33, 52, 53. However, screening withlow-dose spiral CT has been extensively debated because of the controversies mentioned previously 5–8, 52–54. In addition, initial data showed that >50% of individuals had one or more noncalcified nodule(s) at baseline CT examination, rendering more diagnostic steps to be taken 7, 8, 53, 54. Compared to low-dose spiral CT, the “false-positive” rate of early detection for SCC of sputum cytology isvery low. Recent data from the University of Colorado Specialized Programs of Research Excellence (SPORE) trial, conducted in COPD (ex)smokers, have shown that 103 (4.2%) incident cancers developed among 2,441 individuals. This is seven times theincidence of the normal USA population 55.

Immunostaining with heterogeneous nuclear ribonuclear protein (hnRNP) A2/B1 56, quantitative image cytometry by looking at malignant associated changes (MAC) using deoxyribonucleic acid (DNA) texture analysis of the cell nuclei 26 and molecular biological markers 1, 24–29 have revealed changes far in advance and below the detection threshold of morphological classification. Molecular abnormalities may already present prior to morphological changes, therefore, individuals with normal sputum cytology cannot be excluded when studying carcinogenesis. The costs of a screening programme may increase because a larger proportion of COPD (ex)smokers has to be followed. The dynamism of clonal cell changes has tobe studied further. Genetic changes may also be inherent to senescence and clonal dynamism, due to continuous exposure of bronchial mucosa to irritants inducing nonspecific inflammatory dynamic changes. These changes have to be studied diligently for a prolonged period before concluding that they are related to carcinogenesis per se 27, 28, 57. Currently, not a single set of biomarkers has emerged and can beused for accurate prediction of the development oflung cancer in any particular individual. Early molecular changes have been studied, in terms of fieldcarcinogenesis, by looking at “normal” specimens collected from remote areas where the primary tumour was located 27, 28, 58, 59. Changes were detected in about 50–60% of individuals, underscoring the limited knowledge about the dynamism of bronchial mucosa. Therefore, the value of biomarkers hasto be tested prospectively in a larger cohort, comparing normal individuals, COPD, recurrent bronchitis, interstitial lung disease and those who, albeit analysed in retrospect, will develop lung cancer.Studies on archived sputum have shown sensitivities of MAC score (DNA texture analysis) and immunostaining with hnRNP A2/B1 to be 78 and91%, respectively 26, 56. The currently ongoing prospective study of Yunnan tin miners using immunostaining with hnRNP A2/B1 has again shown 80% sensitivity in those with lung cancer primaries and 78% sensitivity to predict those developing lung cancer 56. More importantly, both MAC and immunostaining hnRNP A2/B1 seemed to result in asignificant time shift forward, years in advance beforelung cancer emerged. Semi-automated analysis of nuclear features for MAC has a sensitivity of 75%and a specificity of 90% 26, 60. This allows unattended inspection of 50 slides daily. MAC DNA texture analysis is obviously consistent, excluding intra- and interindividual variability of the pathologists for classifying the degree of preneoplasia. Suspicious cells can be re-examined by the cytopathologist, as the exact co-ordinates of all cells can be digitally stored in a computer. It remains to be seen whether the use of biomarkers, immunostaining with hnRNP A2/B1 and automated sputum cytometry may reduce the range of subjective morphology classification andlead to a significant time shift forward in diagnosis. Currently, the large number of COPD (ex)smokers forms a major hurdle for the realisation of a cost-effective screening programme.

Radiological examinations

In previous trials, chest radiographs have been shown to be insensitive for detecting early parenchymal abnormalities and central airway tumours. Lesions <1 cm in 70% of the cases were missed causing a diagnostic delay of >1 yr 61, 62. Many parenchymal abnormalities were missed when analysed in retrospect. Hence, the use of low-dose spiral CT for early detection obtained a significant stage shift for stage I-II adenocarcinoma 33, 53. This issueremains, however, controversial 5–8.

High-resolution CT improves staging accuracy prior to intraluminal bronchoscopic treatment in central early stage SCC 63. One can exclude bronchial wall abnormalities indicating deeper tumour infiltration. The role of virtual bronchoscopy herein may be limited because of its finite spatial resolution to visualise minute mucosal abnormalities 64. Indeed, small patches of 90,000 cells may prove to be a great challenge for accurate radiological imaging 18, 19, 28. The role of multi-sclice CT herein has notbeen investigated.

Bronchoscopy and autofluorescence bronchoscopy

For SCC, cumulative genetic damages seem to correspond with the different grade of morphological changes, from low to high grade dysplasia (HGD) 1,22, 24–29. Morphologically, normal mucosa may already contain genetic abnormalities 27, 28. Stepwise histological changes are believed to occur from normal through the various grades of squamous metaplasia, mild, moderate and severe dysplasia, and CIS before the lesion becomes micro-invasive 18–21.

This seems compatible with the results of Auerbach and co-workers 18, 20, who investigated the relationship between smoking and changes in bronchial epithelium. Post-mortem step sections of the entire tracheobronchial tree in 402 male patients were obtained, with up to 208 sections per individual. The true presence of CIS in particular was shown to correlate with smoking history and the presence of cancer. CIS lesions were found in 23 (2.4%) of the 956sections, showing atypical cells in the entire specimen, with an average depth of only five cells (range 4–38). In early SCC, 61% of the lesions had depth abnormalities of ≥6 cell layers. Therefore, early SCC is indeed <1 mm thick. Nonsmokers did not appear to harbour any CIS. CIS was present in 75% (27 of 36) of the subjects who smoked ≥2 packs·day⁻¹ and in 83% (52 of 63) of lung cancer subjects. The presence of multiple CISs strongly supports field carcinogenesis, which was found in 11.4% of smokers smoking ≥2 packs·day⁻¹ and in 15% of lung cancer patients.

However, the exact location of CIS using conventional FB could only be determined in 29% of the cases 13. Two-thirds of these ROLC were shown tobe a few mm thick. Sato et al. 65 reported that 527FB sessions were required to ultimately pinpoint the lesions in 180 patients with 200 occult cancers 65. An average of three lengthy ∼45 min FB sessions wascarried out, with separate brushings of all segments. In retrospect, 175 lesions were located proximal to the subsegmental bronchi, thus, within the visible range of the FB. Initial false-negative findings led to an average delay of 29.2 months.

Data clearly show that the lack of any instrument tolocate these lesions will be counterproductive to theconcept of stage shift. The accessibility of the entire bronchial tree is limited by the size of the bronchoscope and the use of an ultrathin bronchoscope for inspecting subsegmental branches up to thetenth generation may only increase the burden ofbronchoscopic examination 66.

Laser induced fluorescence endoscope (AFB-LIFE®; Xillix, Richmond, BC, Canada) uses a helium-cadmium laser of λ 442 nm for tissue excitation 67, 68. The emission spectrum is captured by two sensitive charged coupled device (CCD) cameras and processed through a fluorescence collection sensor and an optical multichannel analyser. Real-time digitised images are acquired by using ratios of red to green (λ 630 nm to λ 520 nm) fluorescence emissions for correcting distance and movement during examinations. Special imaging algorithms have been generated, based on the data of in-vivo and in-vitro analysis, and Monte-Carlo simulation modelling 69. Digitised images reflect the in-vivo fluorescence signals collected, exploiting the differences in autofluorescence characteristics between suspicious (HGD and CIS) and normal bronchial mucosa. The suspicious lesions appear red-brownish while normal mucosa appears green. Other systems use the fluorescence-reflectance system in their imaging algorithms, such as the D-light system (Storz®; Karl Storz, Tüttlingen, Germany). A noncoherent ultraviolet to blue 300 W Xenon filtered lamp (λ 380–460 nm) is used to excite abroad emission spectra of the different chromophores in tissue 70. In combination with some reflectance of blue light, the D-light produces bluish images for normal areas and darker images for HGD and CIS. The SAFE® 1000 autofluorescence system (Pentax, Asahi Optical, Tokyo, Japan) also uses a Xenon-lamp (λ 420–480 nm) and the camera contains a fluorescence filter as well as an image intensifier 71. Also soon, new fluorescence-reflectance imaging systems (ONCO-LIFE®, Xillix; Wolf®, Knittlingen, Germany) will enter clinical trials, with the aim to reduce equipment costs and to ease theautoflouresence bronchoscopy (AFB) procedure. Exploiting both quantitative and qualitative imaging algorithms may improve accuracy in differentiating true pre-invasive lesions from non-specific lesions, thusnonmalignant mucosal abnormalities.

Photodynamic diagnosis, using photosensitisers, has been used to enhance image contrast 70. The relative selectivity of the photosensitising molecules for malignant tissue and the higher photon yield may abolish the need for expensive laser and ultrasensitive CCD cameras. However, the lack of drug selectivity, complex pharmacokinetics and skin photosensitivity raise issues as to whether this will be cost-effective andpractical to be implemented for screening at large.McWilliams et al. 72 reported that the use ofdigitised fluorescence ratio, i.e. quantitative imaging projected on the monitor screen, increased CIS detection rate to the full 100% score.

AFB-LIFE programmes have so far included the largest number of individuals and thousands of lesions have been collected 73–75. A panel of pathologists have reviewed all biopsy specimens for the proper World Health Organization (WHO)/International Association Study Lung Cancer (IASLC) classification, as intra- and interindividual variability of histology reporting is an important determinant of the results. AFB-LIFE in addition to the conventional method of FB has been shown to be 2–4 fold more accurate to detect pre-invasive lesions (80–100% range) 75. The sequence of examinations and the individual bronchoscopist did not affect the results 76. Unforeseen occult synchronous lesions were also found in ∼10% of the individuals with lung cancer primaries 77. AFB-LIFE significantly improved thedetection rate for angiogenic squamous dysplasia, an entity believed to be genetically important in the process of carcinogenesis 76, 78. The additional useof AFB-LIFE for accurate staging helped define the best treatment approach in addition to high-resolution CT, as only 25% of ROLCs were truly occult and amenable for IBT 79. The use of AFB extended the duration of the FB session by<15 min 80. Nevertheless, sensitivity, specificity and accuracy remain a relative issue, strongly dependent on the population studied. Smoking history 75, sex distribution (relative paucity of preneoplastic lesions in females 81), the level of personal expertise of the bronchoscopist 68, 73–75 and the intra- and interindividual variability of histology reporting 82–84 are all factors to be taken into account.

It has been commonly accepted that the WHO/IASLC morphology criteria is the gold standard. However, even among experienced pathologists, full agreement of classifying malignancies may only reacha κ value of 0.73 82. The reproducibility for classifying pre-invasive lesions showed intra-observer agreement of 0.71 and interobserver agreement was only 0.55 83. In classifying CIS in the current authors' initial study period, only <50% were confirmed by the panel to have a CIS 84. Intra- and interobserver variability in histology reports are easilyoverlooked and should be accounted for in analysing data of carcinogenesis studies, especially in looking at specific biomarkers that accompany these early morphological changes 1, 24–28. By using the internet for more accurate morphological classification, a κ value of 0.9 could be obtained 85.

Unfortunately, biopsy sampling using AFB-LIFE has been shown to remove the entire patch of clonal cells 28, 29. Many lesions were <1.5 mm in diameter and ∼50% of these lesions were shown to be smaller than the size of the flexible biopsy forceps. Twenty-seven of the 69 paired biopsies obtained at 6-month intervals showed one or more molecular changes in the initial biopsy specimens. However, 86% had no abnormality after rebiopsy and 24% lacked the initial changes found after repeat biopsy 29. Therefore, the natural history of some lesions cannot be studied because of complete mechanical removal at baseline examination. Studying the natural history of minute pre-invasive lesions, especially at the earlier pre-invasive stages may become virtually impossible.

Spectroscopic images obtain the entire spectral images of the target tissue within a wavelength of interest 86, 87. Optical coherence tomography (OCT) can record two-dimensional, cross-sectional images ofcell layers 88. Confocal micro-endoscopy provides cellular images to be used in studying carcinogenesis at the subcellular level 89. By suppressing images that are out of focus, optical sectioning and three-dimensional reconstruction images can provide superimposed pictures on a microstructural level. Newdevelopments in optical biopsy techniques such as OCT and confocal micro-endoscopy may help intheunderstanding of the natural history of pre-invasive lesions, without the necessity of taking biopsy. However, until the exact algorithms for optical biopsy can be generated, some comparative studies using the WHO/IASLC morphological classification may be required.

The initial false-positive findings using AFB (suspicious fluorescence image but normal histology classification) have to be studied longitudinally. Genetic damages at multiple chromosomal sites (e.g. 3p14, 9p21, p16, etc.) have been reported in morphologically normal and minimally altered bronchial specimens. Increasing percentages of abnormalities were found in current smokers and subjects with a higher metaplasia index 27. Park et al. 28 showed the presence of (multifocal) molecular abnormalities (3p,9p, 17q regions) in morphologically normal and pre-invasive lesions. Increasingly more changes from normal towards dysplastic foci were found 28. In 68% (13 of 19) of the patients the specimens showed atleast one abnormal focus among the seven chromosomal regions analysed. Heterogeneity was seen in thechanges of molecular patterns and they were notnecessarily identical with those in tumour clones. A more homogeneous allele-specific loss was found inmaterials sampled with the AFB-LIFE in their previous study. This suggests that abnormal fluorescence characteristics may represent a more malignant cohort of lesions than in samples collected by random biopsy alone. This seems compatible with therecent interpretation of angiogenic squamous dysplasia 78.

The availability of AFB-LIFE allows for longitudinal study, to exclude sampling bias, in studying the natural history of pre-invasive lesions. Bota et al. 90 recently reported a 6.1% progression rate among 147 dysplastic lesions, irrespective of whether individuals ceased smoking. Only 2.5% of the 359 low-grade lesions progressed and only did so in the presence of a HGD. In 53 patients without HGD orinvasive cancer, none of the low-grade lesions showed progression.

The current authors recently analysed the natural history of pre-invasive lesions 91. Twelve per cent ofthe lesions progressed toward CIS or micro-invasive SCC. The spontaneous regression rate was 88% and no stepwise pattern of morphological changes could be determined. The morphological changes over time were very dynamic in a nonstepwise manner. This was true for all lesions, whether they progressed, regressed or remained stable during follow-up. Progressive lesions were more likely in patients with COPD and after previous primaries. Thedynamism of each lesion during follow-up makes it rather difficult to predict any outcome by using theinitial morphological classification as a starting point. Therefore, the finding of any metaplasia or dysplasia in a transversal study cannot always be reliably used to predict the likelihood for cancer development in any particular individual. The current sampling method of bronchial tissues may contain many nonspecific clonal cell changes and disturb the natural history 57. Nonspecific clonal polymorphisms in the samples collected in a transversal or longitudinal study may generate a high background noise for biomolecular studies, as >80% of the specimens regressed spontaneously. From the data of bothlongitudinal studies 90, 91, the progression rate to malignancy of 6–12% seems compatible with the 10% rate lung cancer development in the population of smokers 17–20. Recent data by Lam et al. 92has shown that in the placebo arm of a chemopreventive agent (anethole dithiolethione) study in smokers, the rate of progression of dysplastic lesions was in the order of 17%.

A follow-up study in individuals who ultimately developed CIS has been conducted with 3–4 monthly repeat AFB-LIFE, showing that all CIS ultimately progressed to SCC 93. CIS seems to be the point of no return, therefore individuals with CIS should be treated promptly. The natural history of preneoplastic lesions have to be further investigated in combination with molecular genetics in a prospective fashion. Optical biopsy techniques may offer a solution to prevent tissue disturbance by taking biopsy, which may disrupt the natural history of these pre-invasive lesions.

How to treat early stage lesions

The most important step prior to any treatment isaccurate staging. “Early stage” SCC should be assessed with regard to tumour margins and depth oftumour invasion in the bronchial wall 13, 30, 31,45. Data of endobronchial ultrasonography have been consistent in terms of its spatial resolution, but itis somewhat early to definitely conclude its value practically 94, 95. The role of PET-scan for accurate staging of intraluminal tumour and to exclude nodal disease also warrants further investigation 96.

The importance of tumour microscopy and its nodal status can be reviewed from previous surgical series 34–37, 97–101. Early detection and surgical resection in previous RCT achieved a 5-yr survival rate of 55 versus 26% in the control arm and the p-value was 0.032 4. It is important to realise that a subgroup of patients with early stage SCC harbour minute superficial and intraluminal lesions. Many phase II studies have dealt with the efficacy of IBT 30, 31, 45, 90, 93, 101–103. Ideally, phase III studies comparing surgery versus various IBT modalities versus a no treatment arm are required. However, the relative paucity of early stage SCC, as many are detected by chance, makes the realisation of such a study difficult. Some individuals are considered poor surgical candidates (e.g. those with COPD, poor cardiovascular status, previously resected syn- and metachronous tumours). IBT is an alternative for surgery and is a parenchyma-conserving technique, provided staging is accurate to stage N0 tumour 45. The choice of IBT modality is of less importance providing local staging has been accurate 101, 102.

Photodynamic therapy (PDT) uses injected photosensitisers, combined with laser illumination of the tumour area to obtain selective necrosis and has therefore been most popular for treating early stage SCC 30, 45, 103. However, lack of selectivity for photosensitisers 87 and scar formation 102–104 donot support this theoretical advantage. All IBT modalities seem to be equally effective 31, 101. Twenty-two out of 39 patients with early stage SCC were spared surgery after initial PDT treatment 45 and any IBT modality prior to surgery may allow forless extensive surgical resections 105–107. IBT asthe initial treatment strategy for intraluminal SCC is a prudent approach to preserve healthy lung tissueand quality of life 45, 79, 93. In 32 patients considered nonsurgical candidates with central early stage SCC, the current authors' data showed that IBT could obtain excellent local control. Follow-up has been 2–10 yrs (median 5 yrs). Half of the patients arealive, and eight died due to recurrences of previous lung cancer primaries, although unrelated to theROLC treatment. The remaining patients died ofnonlung cancer-related causes 108.

The costs of early intervention

The present authors have calculated the cost of AFB-LIFE bronchoscopy for early intervention in atarget population consisting of individuals with previously resected stage I–II NSCLC, ear nose and throat cancers, and positive sputum cytology. Twenty-one lesions progressed to cancer (171 dysplastic lesions and 429 bronchoscopies). The tariff of one bronchoscopy examination in the Netherlands is 81 Euro. The endoscopic cost per cancer lesion found was 1,653 Euro. Treatment with electrocautery costs 380 Euro 101. Thus, early bronchoscopic intervention per early stage SCC costs 2,033 Euro.

Longer follow-up is needed to definitely show the cost-effectiveness of early intervention in the “population at risk”. From the bronchoscopist's point of view,a more pro-active role is warranted as early diagnosis, accurate staging and intraluminal bronchoscopic treatment may provide the best chance for curewhen lung cancer is found at the carcinoma insitustage. However, squamous cell cancer may take10 yrs to develop and repeat diagnostic steps during follow-up increases costs. All controversial aspects mentioned previously should determine the exact value of screening, as increasingly more individuals with poor general condition can be treated usingminimal invasive treatment such as bronchoscopic electrocautery 108. Curative local treatment by intraluminal bronchoscopic treatment increases therisk of these particular individuals to die later from nonlung cancer-related diseases, hence the controversy about treating pseudo-diseases.